Introduction: Raison d’être

ggstatsplot is an extension of ggplot2 package for creating graphics with details from statistical tests included in the plots themselves and targeted primarily at behavioral sciences community to provide a one-line code to produce information-rich plots. In a typical exploratory data analysis workflow, data visualization and statistical modeling are two different phases: visualization informs modeling, and modeling in its turn can suggest a different visualization method, and so on and so forth. The central idea of ggstatsplot is simple: combine these two phases into one in the form of graphics with statistical details, which makes data exploration simpler and faster.

But why would combining statistical analysis with data visualization be helpful? We list few reasons below-

ggstatsplot at a glance

It produces a limited kinds of ready-made plots for the supported analyses:

Function Plot Description
ggbetweenstats violin plots for comparisons between groups/conditions
ggwithinstats violin plots for comparisons within groups/conditions
gghistostats histograms for distribution about numeric variable
ggdotplotstats dot plots/charts for distribution about labeled numeric variable
ggpiestats pie charts for categorical data
ggbarstats bar charts for categorical data
ggscatterstats scatterplots for correlations between two variables
ggcorrmat correlation matrices for correlations between multiple variables
ggcoefstats dot-and-whisker plots for regression models

In addition to these basic plots, ggstatsplot also provides grouped_ versions (see below) that makes it easy to repeat the same analysis for any grouping variable.

Summary of types of statistical analyses

Most functions share a type (of test) argument that is helpful to specify the type of statistical analysis:

The table below summarizes all the different types of analyses currently supported in this package-

Functions Description Parametric Non-parametric Robust Bayes Factor
ggbetweenstats Between group/condition comparisons
ggwithinstats Within group/condition comparisons
gghistostats, ggdotplotstats Distribution of a numeric variable
ggcorrmat Correlation matrix
ggscatterstats Correlation between two variables
ggpiestats, ggbarstats Association between categorical variables NA NA
ggpiestats, ggbarstats Equal proportions for categorical variable levels NA NA
ggcoefstats Regression model coefficients

Graphic design principles

Graphical excellence

Graphical excellence consists of communicating complex ideas with clarity and in a way that the viewer understands the greatest number of ideas in a short amount of time all the while not quoting the data out of context. The package follows the principles for graphical integrity ((???)):

  • The physical representation of numbers is proportional to the numerical quantities they represent (e.g., Figure-1 and Figure-2 show how means (in ggbetweenstats) or percentages (ggpiestats) are proportional to the vertical distance or the area, respectively).

  • All important events in the data have clear, detailed, and thorough labeling (e.g., Figure-1 plot shows how ggbetweenstats labels means, sample size information, outliers, and pairwise comparisons; same can be appreciated for ggpiestats in Figure-2 and gghistostats in Figure-5). Note that data labels in the data region are designed in a way that they don’t interfere with our ability to assess the overall pattern of the data (Cleveland, 1985; p.44-45). This is achieved by using ggrepel package to place labels in a way that reduces their visual prominence.

  • None of the plots have design variation (e.g., abrupt change in scales) over the surface of a same graphic because this can lead to a false impression about variation in data.

  • The number of information-carrying dimensions never exceed the number of dimensions in the data (e.g., using area to show one-dimensional data).

  • All plots are designed to have no chartjunk (like moiré vibrations, fake perspective, dark grid lines, etc.) (???)(???)(Tufte, 2001, Chapter 5).

There are some instances where ggstatsplot graphs don’t follow principles of clean graphics, as formulated in the Tufte theory of data graphics ((???), Chapter 4). The theory has four key principles:(???)

  1. Above all else show the data.
  2. Maximize the data-ink ratio.
  3. Erase non-data-ink.
  4. Erase redundant data-ink, within reason.

In particular, default plots in ggstatsplot can sometimes violate one of the principles from 2-4. According to these principles, every bit of ink should have reason for its inclusion in the graphic and should convey some new information to the viewer. If not, such ink should be removed. One instance of this is bilateral symmetry of data measures. For example, in Figure-1, we can see that both the box and violin plots are mirrored, which consumes twice the space in the graphic without adding any new information. But this redundancy is tolerated for the sake of beauty that such symmetrical shapes can bring to the graphic. Even Tufte admits that efficiency is but one consideration in the design of statistical graphics (Tufte, 2001, p. 137). Additionally, these principles were formulated in an era in which computer graphics had yet to revolutionize the ease with which graphics could be produced and thus some of the concerns about minimizing data-ink for easier production of graphics are not as relevant as they were.

Statistical variation

One of the important functions of a plot is to show the variation in the data, which comes in two forms:

  • Measurement noise: In ggstatsplot, the actual variation in measurements is shown by plotting a combination of (jittered) raw data points with a boxplot laid on top (Figure-1) or a histogram (Figure-5). None of the plots, where empirical distribution of the data is concerned, show the sample standard deviation because they are poor at conveying information about limits of the sample and presence of outliers (Cleveland, 1985, p.220).
# for reproducibility
set.seed(123)

# plot
ggstatsplot::gghistostats(
  data = morley,
  x = Speed,
  test.value = 792,
  test.value.line = TRUE,
  xlab = "Speed of light (km/sec, with 299000 subtracted)",
  title = "Figure-5: Distribution of Speed of light",
  caption = "Note: Data collected across 5 experiments (20 measurements each)",
  messages = FALSE
)
**Figure-5**. Distribution of a variable shown using `gghistostats`.

Figure-5. Distribution of a variable shown using gghistostats.

  • Sample-to-sample statistic variation: Although, traditionally, this variation has been shown using the standard error of the mean (SEM) of the statistic, ggstatsplot plots instead use 95% confidence intervals (e.g., Figure-6). This is because the interval formed by error bars correspond to a 68% confidence interval, which is not a particularly interesting interval (Cleveland, 1985, p.222-225).
# for reproducibility
set.seed(123)

# creating model object
mod <- lme4::lmer(
  formula = total.fruits ~ nutrient + rack + (nutrient | popu / gen),
  data = lme4::Arabidopsis
)
#> boundary (singular) fit: see ?isSingular

# plot
ggstatsplot::ggcoefstats(
  x = mod,
  p.kr = FALSE
)
#> Computing p-values via Wald-statistics approximation (treating t as Wald z).
**Figure-6**. Sample-to-sample variation in regression estimates is displayed using confidence intervals in `ggcoefstats`.

Figure-6. Sample-to-sample variation in regression estimates is displayed using confidence intervals in ggcoefstats.

Statistical analysis

As an extension of ggplot2, ggstatsplot has the same expectations about the structure of the data. More specifically,


# creating a new dataset without any NAs in variables of interest
msleep_no_na <-
  dplyr::filter(
    .data = ggplot2::msleep,
    !is.na(sleep_rem), !is.na(awake), !is.na(brainwt), !is.na(bodywt)
  )

# variable names vector
var_names <- c("REM sleep", "time awake", "brain weight", "body weight")

# combining two plots using helper function in `ggstatsplot`
ggstatsplot::combine_plots(
  # plot *without* any NAs
  ggstatsplot::ggcorrmat(
    data = msleep_no_na,
    p.adjust.method = "holm",
    cor.vars = c(sleep_rem, awake:bodywt),
    cor.vars.names = var_names,
    matrix.type = "upper",
    colors = c("#B2182B", "white", "#4D4D4D"),
    title = "Correlalogram for mammals sleep dataset",
    subtitle = "sleep units: hours; weight units: kilograms",
    messages = FALSE
  ),
  # plot *with* NAs
  ggstatsplot::ggcorrmat(
    data = ggplot2::msleep,
    p.adjust.method = "holm",
    cor.vars = c(sleep_rem, awake:bodywt),
    cor.vars.names = var_names,
    matrix.type = "upper",
    colors = c("#B2182B", "white", "#4D4D4D"),
    title = "Correlalogram for mammals sleep dataset",
    subtitle = "sleep units: hours; weight units: kilograms",
    messages = FALSE
  ),
  labels = c("(a)", "(b)"),
  nrow = 1
)
**Figure-4**. `ggstatsplot` functions remove `NA`s from variables of interest and display total sample size *n*, but they can give more nuanced information about sample sizes when *n* differs across tests. For example, `ggcorrmat` will display (a) only one total sample size once when no `NA`s present, but (b) will instead show minimum, median, and maximum sample sizes across all correlation tests when `NA`s are present across correlation variables.

Figure-4. ggstatsplot functions remove NAs from variables of interest and display total sample size n, but they can give more nuanced information about sample sizes when n differs across tests. For example, ggcorrmat will display (a) only one total sample size once when no NAs present, but (b) will instead show minimum, median, and maximum sample sizes across all correlation tests when NAs are present across correlation variables.

Statistical reporting

For all statistical tests reported in the plots, the default template abides by the APA gold standard for statistical reporting. For example, here are results from Yuen’s test for trimmed means (robust t-test):

Template for reporting statistical details

Template for reporting statistical details

Here is a summary table of all the statistical tests currently supported across various functions:

Functions Type Test Effect size 95% CI available?
ggbetweenstats (2 groups) Parametric Student’s and Welch’s t-test Cohen’s d, Hedge’s g
ggbetweenstats (> 2 groups) Parametric Fisher’s and Welch’s one-way ANOVA \[\eta^2, \eta^2_p, \omega^2, \omega^2_p\]
ggbetweenstats (2 groups) Non-parametric Mann-Whitney U-test r
ggbetweenstats (> 2 groups) Non-parametric Kruskal-Wallis Rank Sum Test \[\epsilon^2\]
ggbetweenstats (2 groups) Robust Yuen’s test for trimmed means \[\xi\]
ggbetweenstats (> 2 groups) Robust Heteroscedastic one-way ANOVA for trimmed means \[\xi\]
ggwithinstats (2 groups) Parametric Student’s t-test Cohen’s d, Hedge’s g
ggwithinstats (> 2 groups) Parametric Fisher’s one-way repeated measures ANOVA \[\eta^2_p, \omega^2\]
ggwithinstats (2 groups) Non-parametric Wilcoxon signed-rank test r
ggwithinstats (> 2 groups) Non-parametric Friedman rank sum test \[W_{Kendall}\]
ggwithinstats (2 groups) Robust Yuen’s test on trimmed means for dependent samples \[\xi\]
ggwithinstats (> 2 groups) Robust Heteroscedastic one-way repeated measures ANOVA for trimmed means
ggpiestats and ggbarstats (unpaired) Parametric \[\text{Pearson's}~ \chi^2 ~\text{test}\] Cramér’s V
ggpiestats and ggbarstats (paired) Parametric McNemar’s test Cohen’s g
ggpiestats Parametric One-sample proportion test Cramér’s V
ggscatterstats and ggcorrmat Parametric Pearson’s r r
ggscatterstats and ggcorrmat Non-parametric \[\text{Spearman's}~ \rho\] \[\rho\]
ggscatterstatsand ggcorrmat Robust Percentage bend correlation r
gghistostats and ggdotplotstats Parametric One-sample t-test Cohen’s d, Hedge’s g
gghistostats Non-parametric One-sample Wilcoxon signed rank test r
gghistostats and ggdotplotstats Robust One-sample percentile bootstrap robust estimator
ggcoefstats Parametric Regression models \[\beta\]

Reporting results

The default setting in ggstatsplot is to produce plots with statistical details included. Most often than not, the results are displayed as a subtitle in the plot. Great care has been taken into which details are included in statistical reporting and why.

  1. APA guidelines (APA, 2009) are followed (for the most part) by default:
  • Percentages are displayed with no decimal places (Figure-2).
  • Correlations, t-tests, and \(\chi^2\)-tests are reported with the degrees of freedom in parentheses and the significance level (Figure-2, Figure-3b, Figure-5).
  • ANOVAs are reported with two degrees of freedom and the significance level (Figure-1).
  • Regression results are presented with the unstandardized or standardized estimate (beta), whichever was specified by the user, along with the statistic (depending on the model, this can be a t or z statistic) and the corresponding significance level (Figure-6).
  • With the exception of p-values, most statistics are rounded to two decimal places by default.

  1. Default statistical tests:

  2. Dealing with null results:

  3. Avoiding the “p-value error”:
    The p-value indexes the probability that the researchers have falsely rejected a true null hypothesis (Type I error, i.e.) and can rarely be exactly 0. And yet over 97,000 manuscripts on Google Scholar report the p-value to be p = 0.000 (???), putatively due to relying on default computer outputs. All p-values displayed in ggstatsplot plots avoid this mistake. Anything less than p < 0.001 is displayed as such (e.g, Figure-1). The package deems it unimportant how infinitesimally small the p-values are and, instead, puts emphasis on the effect size magnitudes and their 95% CIs.

Short tutorial on primary functions

Here are examples of the main functions currently supported in ggstatsplot.

ggbetweenstats

This function creates either a violin plot, a box plot, or a mix of two for between-group or between-condition comparisons with results from statistical tests in the subtitle. The simplest function call looks like this-

# loading needed libraries
library(ggstatsplot)

# for reproducibility
set.seed(123)

# plot
ggstatsplot::ggbetweenstats(
  data = iris,
  x = Species,
  y = Sepal.Length,
  messages = FALSE
) + # further modification outside of ggstatsplot
  ggplot2::coord_cartesian(ylim = c(3, 8)) +
  ggplot2::scale_y_continuous(breaks = seq(3, 8, by = 1))

Note that this function returns a ggplot2 object and thus any of the graphics layers can be further modified.

A number of other arguments can be specified to make this plot even more informative or change some of the default options.

library(ggplot2)

# for reproducibility
set.seed(123)

# let's leave out one of the factor levels and see if instead of anova, a t-test will be run
iris2 <- dplyr::filter(.data = iris, Species != "setosa")

# let's change the levels of our factors, a common routine in data analysis
# pipeline, to see if this function respects the new factor levels
iris2$Species <- factor(x = iris2$Species, levels = c("virginica", "versicolor"))

# plot
ggstatsplot::ggbetweenstats(
  data = iris2,
  x = Species,
  y = Sepal.Length,
  notch = TRUE, # show notched box plot
  mean.plotting = TRUE, # whether mean for each group is to be displayed
  mean.ci = TRUE, # whether to display confidence interval for means
  mean.label.size = 2.5, # size of the label for mean
  type = "parametric", # which type of test is to be run
  k = 3, # number of decimal places for statistical results
  outlier.tagging = TRUE, # whether outliers need to be tagged
  outlier.label = Sepal.Width, # variable to be used for the outlier tag
  outlier.label.color = "darkgreen", # changing the color for the text label
  xlab = "Type of Species", # label for the x-axis variable
  ylab = "Attribute: Sepal Length", # label for the y-axis variable
  title = "Dataset: Iris flower data set", # title text for the plot
  ggtheme = ggthemes::theme_fivethirtyeight(), # choosing a different theme
  ggstatsplot.layer = FALSE, # turn off ggstatsplot theme layer
  package = "wesanderson", # package from which color palette is to be taken
  palette = "Darjeeling1", # choosing a different color palette
  messages = FALSE
)

As can be seen from the plot, the function by default returns Bayes Factor for the test. If the null hypothesis can’t be rejected with the null hypothesis significance testing (NHST) approach, the Bayesian approach can help index evidence in favor of the null hypothesis (i.e., \(BF_{01}\)).

By default, natural logarithms are shown because Bayes Factor values can sometimes be pretty large. Having values on logarithmic scale also makes it easy to compare evidence in favor alternative (\(BF_{10}\)) versus null (\(BF_{01}\)) hypotheses (since \(log_{e}(BF_{01}) = - log_{e}(BF_{01})\)).

Additionally, there is also a grouped_ variant of this function that makes it easy to repeat the same operation across a single grouping variable:

# for reproducibility
set.seed(123)

# plot
ggstatsplot::grouped_ggbetweenstats(
  data = dplyr::filter(
    .data = ggstatsplot::movies_long,
    genre %in% c("Action", "Action Comedy", "Action Drama", "Comedy")
  ),
  x = mpaa,
  y = length,
  grouping.var = genre, # grouping variable
  pairwise.comparisons = TRUE, # display significant pairwise comparisons
  pairwise.annotation = "p.value", # how do you want to annotate the pairwise comparisons
  p.adjust.method = "bonferroni", # method for adjusting p-values for multiple comparisons
  conf.level = 0.99, # changing confidence level to 99%
  ggplot.component = list( # adding new components to `ggstatsplot` default
    ggplot2::scale_y_continuous(sec.axis = ggplot2::dup_axis())
  ),
  k = 3,
  title.prefix = "Movie genre",
  caption = substitute(paste(
    italic("Source"),
    ":IMDb (Internet Movie Database)"
  )),
  palette = "default_jama",
  package = "ggsci",
  messages = FALSE,
  nrow = 2,
  title.text = "Differences in movie length by mpaa ratings for different genres"
)

Summary of tests

Following (between-subjects) tests are carried out for each type of analyses-

Type No. of groups Test
Parametric > 2 Fisher’s or Welch’s one-way ANOVA
Non-parametric > 2 Kruskal–Wallis one-way ANOVA
Robust > 2 Heteroscedastic one-way ANOVA for trimmed means
Bayes Factor > 2 Fisher’s ANOVA
Parametric 2 Student’s or Welch’s t-test
Non-parametric 2 Mann–Whitney U test
Robust 2 Yuen’s test for trimmed means
Bayes Factor 2 Student’s t-test

The omnibus effect in one-way ANOVA design can also be followed up with more focal pairwise comparison tests. Here is a summary of multiple pairwise comparison tests supported in ggbetweenstats-

Type Equal variance? Test p-value adjustment?
Parametric No Games-Howell test
Parametric Yes Student’s t-test
Non-parametric No Dwass-Steel-Crichtlow-Fligner test
Robust No Yuen’s trimmed means test
Bayes Factor No
Bayes Factor Yes

For more, see the ggbetweenstats vignette: https://indrajeetpatil.github.io/ggstatsplot/articles/web_only/ggbetweenstats.html

ggwithinstats

ggbetweenstats function has an identical twin function ggwithinstats for repeated measures designs that behaves in the same fashion with a few minor tweaks introduced to properly visualize the repeated measures design. As can be seen from an example below, the only difference between the plot structure is that now the group means are connected by paths to highlight the fact that these data are paired with each other.

# for reproducibility and data
set.seed(123)
library(WRS2)

# plot
ggstatsplot::ggwithinstats(
  data = WRS2::WineTasting,
  x = Wine,
  y = Taste,
  sort = "descending", # ordering groups along the x-axis based on
  sort.fun = median, # values of `y` variable
  pairwise.comparisons = TRUE,
  pairwise.display = "s",
  pairwise.annotation = "p",
  title = "Wine tasting",
  caption = "Data from: `WRS2` R package",
  ggtheme = ggthemes::theme_fivethirtyeight(),
  ggstatsplot.layer = FALSE,
  messages = FALSE
)

As with the ggbetweenstats, this function also has a grouped_ variant that makes repeating the same analysis across a single grouping variable quicker. We will see an example with only repeated measurements-

# common setup
set.seed(123)

# getting data in tidy format
data_bugs <- ggstatsplot::bugs_long %>%
  dplyr::filter(.data = ., region %in% c("Europe", "North America"))

# plot
ggstatsplot::grouped_ggwithinstats(
  data = dplyr::filter(data_bugs, condition %in% c("LDLF", "LDHF")),
  x = condition,
  y = desire,
  xlab = "Condition",
  ylab = "Desire to kill an artrhopod",
  grouping.var = region,
  outlier.tagging = TRUE,
  outlier.label = education,
  ggtheme = ggthemes::theme_fivethirtyeight(),
  ggstatsplot.layer = FALSE,
  messages = FALSE
)

Summary of tests

Following (within-subjects) tests are carried out for each type of analyses-

Type No. of groups Test
Parametric > 2 One-way repeated measures ANOVA
Non-parametric > 2 Friedman test
Robust > 2 Heteroscedastic one-way repeated measures ANOVA for trimmed means
Bayes Factor > 2 One-way repeated measures ANOVA
Parametric 2 Student’s t-test
Non-parametric 2 Wilcoxon signed-rank test
Robust 2 Yuen’s test on trimmed means for dependent samples
Bayes Factor 2 Student’s t-test

The omnibus effect in one-way ANOVA design can also be followed up with more focal pairwise comparison tests. Here is a summary of multiple pairwise comparison tests supported in ggwithinstats-

Type Test p-value adjustment?
Parametric Student’s t-test
Non-parametric Durbin-Conover test
Robust Yuen’s trimmed means test
Bayes Factor

For more, see the ggwithinstats vignette: https://indrajeetpatil.github.io/ggstatsplot/articles/web_only/ggwithinstats.html

ggscatterstats

This function creates a scatterplot with marginal distributions overlaid on the axes (from ggExtra::ggMarginal) and results from statistical tests in the subtitle:

ggstatsplot::ggscatterstats(
  data = ggplot2::msleep,
  x = sleep_rem,
  y = awake,
  xlab = "REM sleep (in hours)",
  ylab = "Amount of time spent awake (in hours)",
  title = "Understanding mammalian sleep",
  messages = FALSE
)

The available marginal distributions are-

  • histograms
  • boxplots
  • density
  • violin
  • densigram (density + histogram)

Number of other arguments can be specified to modify this basic plot-

# for reproducibility
set.seed(123)

# plot
ggstatsplot::ggscatterstats(
  data = dplyr::filter(.data = ggstatsplot::movies_long, genre == "Action"),
  x = budget,
  y = rating,
  type = "robust", # type of test that needs to be run
  conf.level = 0.99, # confidence level
  xlab = "Movie budget (in million/ US$)", # label for x axis
  ylab = "IMDB rating", # label for y axis
  label.var = "title", # variable for labeling data points
  label.expression = "rating < 5 & budget > 100", # expression that decides which points to label
  line.color = "yellow", # changing regression line color line
  title = "Movie budget and IMDB rating (action)", # title text for the plot
  caption = expression( # caption text for the plot
    paste(italic("Note"), ": IMDB stands for Internet Movie DataBase")
  ),
  ggtheme = hrbrthemes::theme_ipsum_ps(), # choosing a different theme
  ggstatsplot.layer = FALSE, # turn off ggstatsplot theme layer
  marginal.type = "density", # type of marginal distribution to be displayed
  xfill = "#0072B2", # color fill for x-axis marginal distribution
  yfill = "#009E73", # color fill for y-axis marginal distribution
  xalpha = 0.6, # transparency for x-axis marginal distribution
  yalpha = 0.6, # transparency for y-axis marginal distribution
  centrality.para = "median", # central tendency lines to be displayed
  messages = FALSE # turn off messages and notes
)

Additionally, there is also a grouped_ variant of this function that makes it easy to repeat the same operation across a single grouping variable. Also, note that, as opposed to the other functions, this function does not return a ggplot object and any modification you want to make can be made in advance using ggplot.component argument (available for all functions, but especially useful for this particular function):

# for reproducibility
set.seed(123)

# plot
ggstatsplot::grouped_ggscatterstats(
  data = dplyr::filter(
    .data = ggstatsplot::movies_long,
    genre %in% c("Action", "Action Comedy", "Action Drama", "Comedy")
  ),
  x = rating,
  y = length,
  label.var = title,
  label.expression = length > 200,
  conf.level = 0.99,
  k = 3, # no. of decimal places in the results
  xfill = "#E69F00",
  yfill = "#8b3058",
  xlab = "IMDB rating",
  grouping.var = genre, # grouping variable
  title.prefix = "Movie genre",
  ggtheme = ggplot2::theme_grey(),
  ggplot.component = list(
    ggplot2::scale_x_continuous(breaks = seq(2, 9, 1), limits = (c(2, 9)))
  ),
  messages = FALSE,
  nrow = 2,
  title.text = "Relationship between movie length by IMDB ratings for different genres"
)

Summary of tests

Following tests are carried out for each type of analyses. Additionally, the correlation coefficients (and their confidence intervals) are used as effect sizes-

Type Test CI?
Parametric Pearson’s correlation coefficient
Non-parametric Spearman’s rank correlation coefficient
Robust Percentage bend correlation coefficient
Bayes Factor Pearson’s correlation coefficient

For more, see the ggscatterstats vignette: https://indrajeetpatil.github.io/ggstatsplot/articles/web_only/ggscatterstats.html

ggpiestats

This function creates a pie chart for categorical or nominal variables with results from contingency table analysis (Pearson’s \(\chi^2\) test for between-subjects design and McNemar’s \(\chi^2\) test for within-subjects design) included in the subtitle of the plot. If only one categorical variable is entered, results from one-sample proportion test (i.e., a \(\chi^2\) goodness of fit/gof test) will be displayed as a subtitle.

Here is an example of a case where the theoretical question is about proportions for different levels of a single nominal variable:

# for reproducibility
set.seed(123)

# plot
ggstatsplot::ggpiestats(
  data = ggplot2::msleep,
  x = vore,
  title = "Composition of vore types among mammals",
  messages = FALSE
)

This function can also be used to study an interaction between two categorical variables:

# for reproducibility
set.seed(123)

# plot
ggstatsplot::ggpiestats(
  data = mtcars,
  x = am,
  y = cyl,
  conf.level = 0.99, # confidence interval for effect size measure
  title = "Dataset: Motor Trend Car Road Tests", # title for the plot
  stat.title = "interaction: ", # title for the results
  legend.title = "Transmission", # title for the legend
  factor.levels = c("1 = manual", "0 = automatic"), # renaming the factor level names (`x`)
  facet.wrap.name = "No. of cylinders", # name for the facetting variable
  slice.label = "counts", # show counts data instead of percentages
  package = "ggsci", # package from which color palette is to be taken
  palette = "default_jama", # choosing a different color palette
  caption = substitute( # text for the caption
    paste(italic("Source"), ": 1974 Motor Trend US magazine")
  ),
  messages = FALSE # turn off messages and notes
)

In case of repeated measures designs, setting paired = TRUE will produce results from McNemar’s \(\chi^2\) test-

# for reproducibility
set.seed(123)

# data
survey.data <- data.frame(
  `1st survey` = c("Approve", "Approve", "Disapprove", "Disapprove"),
  `2nd survey` = c("Approve", "Disapprove", "Approve", "Disapprove"),
  `Counts` = c(794, 150, 86, 570),
  check.names = FALSE
)

# plot
ggstatsplot::ggpiestats(
  data = survey.data,
  x = `1st survey`,
  y = `2nd survey`,
  counts = Counts,
  paired = TRUE, # within-subjects design
  conf.level = 0.99, # confidence interval for effect size measure
  package = "wesanderson",
  palette = "Royal1"
)
#> Note: 99% CI for effect size estimate was computed with 100 bootstrap samples.
#> Note: Results from one-sample proportion tests for each level of the variable
#> 2nd survey testing for equal proportions of the variable 1st survey.
#> # A tibble: 2 x 10
#>   `2nd survey` counts  perc N     Approve Disapprove `Chi-squared`    df
#>   <fct>         <int> <dbl> <chr> <chr>   <chr>              <dbl> <dbl>
#> 1 Disapprove      720   45  (n =~ 20.83%  79.17%              245      1
#> 2 Approve         880   55. (n =~ 90.23%  9.77%               570.     1
#> # ... with 2 more variables: `p-value` <dbl>, significance <chr>

Note that when a two-way table is present (i.e., when both x and y arguments are specified), p-values for results from one-sample proportion tests are displayed in each facet in the form of asterisks with the following convention:

  • \(***\): \(p < 0.001\)
  • \(**\): \(p < 0.01\)
  • \(*\): \(p < 0.05\)
  • \(ns\): \(p > 0.05\)

Additionally, there is also a grouped_ variant of this function that makes it easy to repeat the same operation across a single grouping variable:

# for reproducibility
set.seed(123)

# plot
ggstatsplot::grouped_ggpiestats(
  dplyr::filter(
    .data = ggstatsplot::movies_long,
    genre %in% c("Action", "Action Comedy", "Action Drama", "Comedy")
  ),
  x = mpaa,
  grouping.var = genre, # grouping variable
  title.prefix = "Movie genre", # prefix for the facetted title
  label.text.size = 3, # text size for slice labels
  slice.label = "both", # show both counts and percentage data
  perc.k = 1, # no. of decimal places for percentages
  palette = "brightPastel",
  package = "quickpalette",
  messages = FALSE,
  nrow = 2,
  title.text = "Composition of MPAA ratings for different genres"
)

Summary of tests

Following tests are carried out for each type of analyses-

Type of data Design Test
Unpaired \(n \times p\) contingency table Pearson’s \(\chi^{2}\) test
Paired \(n \times p\) contingency table McNemar’s \(\chi^{2}\) test
Frequency \(n \times 1\) contingency table Goodness of fit (\(\chi^{2}\))

Following effect sizes (and confidence intervals/CI) are available for each type of test-

Type Effect size CI?
Pearson’s chi-squared test Cramér’s V
McNemar’s test Cohen’s g
Goodness of fit Cramér’s V

For more, see the ggpiestats vignette: https://indrajeetpatil.github.io/ggstatsplot/articles/web_only/ggpiestats.html

ggbarstats

In case you are not a fan of pie charts (for very good reasons), you can alternatively use ggbarstats function which has a similar syntax-

# for reproducibility
set.seed(123)

# plot
ggstatsplot::ggbarstats(
  data = ggstatsplot::movies_long,
  x = mpaa,
  y = genre,
  sampling.plan = "jointMulti",
  title = "MPAA Ratings by Genre",
  xlab = "movie genre",
  perc.k = 1,
  x.axis.orientation = "slant",
  ggtheme = hrbrthemes::theme_modern_rc(),
  ggstatsplot.layer = FALSE,
  ggplot.component = ggplot2::theme(axis.text.x = ggplot2::element_text(face = "italic")),
  palette = "Set2",
  messages = FALSE
)

And, needless to say, there is also a grouped_ variant of this function-

# setup
set.seed(123)

# smaller dataset
df <- dplyr::filter(
  .data = forcats::gss_cat,
  race %in% c("Black", "White"),
  relig %in% c("Protestant", "Catholic", "None"),
  !partyid %in% c("No answer", "Don't know", "Other party")
)

# plot
ggstatsplot::grouped_ggbarstats(
  data = df,
  x = relig,
  y = partyid,
  grouping.var = race,
  title.prefix = "Race",
  xlab = "Party affiliation",
  ggtheme = ggthemes::theme_tufte(base_size = 12),
  ggstatsplot.layer = FALSE,
  messages = FALSE,
  title.text = "Race, religion, and political affiliation",
  nrow = 2
)

Summary of tests

This is identical to the ggpiestats function summary of tests.

gghistostats

In case you would like to see the distribution of a single variable and check if it is significantly different from a specified value with a one sample test, this function will let you do that.

# for reproducibility
set.seed(123)

# plot
ggstatsplot::gghistostats(
  data = iris, # dataframe from which variable is to be taken
  x = Sepal.Length, # numeric variable whose distribution is of interest
  title = "Distribution of Iris sepal length", # title for the plot
  caption = substitute(paste(italic("Source:", "Ronald Fisher's Iris data set"))),
  type = "parametric", # one sample t-test
  conf.level = 0.99, # changing confidence level for effect size
  bar.measure = "mix", # what does the bar length denote
  test.value = 5, # default value is 0
  test.value.line = TRUE, # display a vertical line at test value
  test.value.color = "#0072B2", # color for the line for test value
  centrality.para = "mean", # which measure of central tendency is to be plotted
  centrality.color = "darkred", # decides color for central tendency line
  binwidth = 0.10, # binwidth value (experiment)
  bf.prior = 0.8, # prior width for computing bayes factor
  messages = FALSE, # turn off the messages
  ggtheme = hrbrthemes::theme_ipsum_tw(), # choosing a different theme
  ggstatsplot.layer = FALSE # turn off ggstatsplot theme layer
)

Additionally, there is also a grouped_ variant of this function that makes it easy to repeat the same operation across a single grouping variable:

# for reproducibility
set.seed(123)

# plot
ggstatsplot::grouped_gghistostats(
  data = dplyr::filter(
    .data = ggstatsplot::movies_long,
    genre %in% c("Action", "Action Comedy", "Action Drama", "Comedy")
  ),
  x = budget,
  xlab = "Movies budget (in million US$)",
  type = "robust", # use robust location measure
  grouping.var = genre, # grouping variable
  normal.curve = TRUE, # superimpose a normal distribution curve
  normal.curve.color = "red",
  title.prefix = "Movie genre",
  ggtheme = ggthemes::theme_tufte(),
  ggplot.component = list( # modify the defaults from `ggstatsplot` for each plot
    ggplot2::scale_x_continuous(breaks = seq(0, 200, 50), limits = (c(0, 200)))
  ),
  messages = FALSE,
  nrow = 2,
  title.text = "Movies budgets for different genres"
)

Summary of tests

Following tests are carried out for each type of analyses-

Type Test
Parametric One-sample Student’s t-test
Non-parametric One-sample Wilcoxon test
Robust One-sample percentile bootstrap
Bayes Factor One-sample Student’s t-test

Following effect sizes (and confidence intervals/CI) are available for each type of test-

Type Effect size CI?
Parametric Cohen’s d, Hedge’s g (central-and noncentral-t distribution based)
Non-parametric r (computed as \(Z/\sqrt{N}\))
Robust \(M_{robust}\) (Robust location measure)
Bayes Factor

For more, including information about the variant of this function grouped_gghistostats, see the gghistostats vignette: https://indrajeetpatil.github.io/ggstatsplot/articles/web_only/gghistostats.html

ggdotplotstats

This function is similar to gghistostats, but is intended to be used when the numeric variable also has a label.

# for reproducibility
set.seed(123)

# plot
ggdotplotstats(
  data = dplyr::filter(.data = gapminder::gapminder, continent == "Asia"),
  y = country,
  x = lifeExp,
  test.value = 55,
  test.value.line = TRUE,
  test.line.labeller = TRUE,
  test.value.color = "red",
  centrality.para = "median",
  centrality.k = 0,
  title = "Distribution of life expectancy in Asian continent",
  xlab = "Life expectancy",
  messages = FALSE,
  caption = substitute(
    paste(
      italic("Source"),
      ": Gapminder dataset from https://www.gapminder.org/"
    )
  )
)

As with the rest of the functions in this package, there is also a grouped_ variant of this function to facilitate looping the same operation for all levels of a single grouping variable.

# for reproducibility
set.seed(123)

# removing factor level with very few no. of observations
df <- dplyr::filter(.data = ggplot2::mpg, cyl %in% c("4", "6"))

# plot
ggstatsplot::grouped_ggdotplotstats(
  data = df,
  x = cty,
  y = manufacturer,
  xlab = "city miles per gallon",
  ylab = "car manufacturer",
  type = "nonparametric", # non-parametric test
  grouping.var = cyl, # grouping variable
  test.value = 15.5,
  title.prefix = "cylinder count",
  point.color = "red",
  point.size = 5,
  point.shape = 13,
  test.value.line = TRUE,
  ggtheme = ggthemes::theme_par(),
  messages = FALSE,
  title.text = "Fuel economy data"
)

Summary of tests

This is identical to summary of tests for gghistostats.

ggcorrmat

ggcorrmat makes a correlalogram (a matrix of correlation coefficients) with minimal amount of code. Just sticking to the defaults itself produces publication-ready correlation matrices. But, for the sake of exploring the available options, let’s change some of the defaults. For example, multiple aesthetics-related arguments can be modified to change the appearance of the correlation matrix.

# for reproducibility
set.seed(123)

# as a default this function outputs a correlalogram plot
ggstatsplot::ggcorrmat(
  data = ggplot2::msleep,
  corr.method = "robust", # correlation method
  sig.level = 0.001, # threshold of significance
  p.adjust.method = "holm", # p-value adjustment method for multiple comparisons
  cor.vars = c(sleep_rem, awake:bodywt), # a range of variables can be selected
  cor.vars.names = c(
    "REM sleep", # variable names
    "time awake",
    "brain weight",
    "body weight"
  ),
  matrix.type = "upper", # type of visualization matrix
  colors = c("#B2182B", "white", "#4D4D4D"),
  title = "Correlalogram for mammals sleep dataset",
  subtitle = "sleep units: hours; weight units: kilograms"
)

Note that if there are NAs present in the selected variables, the legend will display minimum, median, and maximum number of pairs used for correlation tests.

There is a grouped_ variant of this function that makes it easy to repeat the same operation across a single grouping variable:

# for reproducibility
set.seed(123)

# plot
# let's use only 50% of the data to speed up the process
ggstatsplot::grouped_ggcorrmat(
  data = dplyr::filter(
    .data = ggstatsplot::movies_long,
    genre %in% c("Action", "Action Comedy", "Action Drama", "Comedy")
  ),
  cor.vars = length:votes,
  corr.method = "np",
  colors = c("#cbac43", "white", "#550000"),
  grouping.var = genre, # grouping variable
  digits = 3, # number of digits after decimal point
  title.prefix = "Movie genre",
  messages = FALSE,
  nrow = 2
)

Summary of tests

Following tests are carried out for each type of analyses. Additionally, the correlation coefficients (and their confidence intervals) are used as effect sizes-

Type Test CI?
Parametric Pearson’s correlation coefficient
Non-parametric Spearman’s rank correlation coefficient
Robust Percentage bend correlation coefficient
Bayes Factor Pearson’s correlation coefficient

For examples and more information, see the ggcorrmat vignette: https://indrajeetpatil.github.io/ggstatsplot/articles/web_only/ggcorrmat.html

ggcoefstats

ggcoefstats creates a dot-and-whisker plot for regression models. Although the statistical models displayed in the plot may differ based on the class of models being investigated, there are few aspects of the plot that will be invariant across models:

  • The dot-whisker plot contains a dot representing the estimate and their confidence intervals (95% is the default). The estimate can either be effect sizes (for tests that depend on the F statistic) or regression coefficients (for tests with t and z statistic), etc. The function will, by default, display a helpful x-axis label that should clear up what estimates are being displayed. The confidence intervals can sometimes be asymmetric if bootstrapping was used.

  • The caption will always contain diagnostic information, if available, about models that can be useful for model selection: The smaller the Akaike’s Information Criterion (AIC) and the Bayesian Information Criterion (BIC) values, the “better” the model is. Additionally, the higher the log-likelihood value the “better” is the model fit.

  • The output of this function will be a ggplot2 object and, thus, it can be further modified (e.g., change themes, etc.) with ggplot2 functions.

# for reproducibility
set.seed(123)

# model
mod <- stats::lm(formula = mpg ~ am * cyl, data = mtcars)

# plot
ggstatsplot::ggcoefstats(x = mod, title = "linear model")

This default plot can be further modified to one’s liking with additional arguments (also, let’s use a robust linear model instead of a simple linear model now):

# for reproducibility
set.seed(123)

# model
mod <- MASS::rlm(formula = mpg ~ am * cyl, data = mtcars)

# plot
ggstatsplot::ggcoefstats(
  x = mod,
  point.color = "red",
  point.shape = 15,
  vline.color = "#CC79A7",
  vline.linetype = "dotdash",
  stats.label.size = 3.5,
  stats.label.color = c("#0072B2", "#D55E00", "darkgreen"),
  title = "Car performance predicted by transmission & cylinder count",
  subtitle = "Source: 1974 Motor Trend US magazine",
  ggtheme = hrbrthemes::theme_ipsum_ps(),
  ggstatsplot.layer = FALSE
) + # further modification with the ggplot2 commands
  ggplot2::scale_y_discrete(labels = c("transmission", "cylinders", "interaction")) +
  ggplot2::labs(x = "regression coefficient", y = NULL)

Most of the regression models that are supported in the broom and broom.mixed packages with tidy and glance methods are also supported by ggcoefstats. For example-

aareg, anova, aov, aovlist, Arima, bigglm, biglm, brmsfit, btergm, cch, clm, clmm, confusionMatrix, coxph, drc, emmGrid, epi.2by2, ergm, felm, fitdistr, glmerMod, glmmTMB, gls, gam, Gam, gamlss, garch, glm, glmmadmb, glmmPQL, glmmTMB, glmRob, glmrob, gmm, ivreg, lm, lm.beta, lmerMod, lmodel2, lmRob, lmrob, mcmc, MCMCglmm, mediate, mjoint, mle2, mlm, multinom, negbin, nlmerMod, nlrq, nls, orcutt, plm, polr, ridgelm, rjags, rlm, rlmerMod, rq, speedglm, speedlm, stanreg, survreg, svyglm, svyolr, svyglm, etc.

Although not shown here, this function can also be used to carry out both frequentist and Bayesian random-effects meta-analysis.

For a more exhaustive account of this function, see the associated vignette- https://indrajeetpatil.github.io/ggstatsplot/articles/web_only/ggcoefstats.html

Working with custom plots

Sometimes you may not like the defaults in a plot produced by ggstatsplot. In such cases, you can use other custom plots (from ggplot2 or other graphics packages in R) and still use ggstatsplot functions to display results from relevant statistical test.

For example, in the following chunk, we will create plot (pirateplot) using yarrr package and use ggstatsplot function for extracting results.

# for reproducibility
set.seed(123)

# loading the needed libraries
library(yarrr)
library(ggstatsplot)

# using `ggstatsplot` to get call with statistical results
stats_results <-
  ggstatsplot::ggbetweenstats(
    data = ChickWeight,
    x = Time,
    y = weight,
    return = "subtitle",
    messages = FALSE
  )

# using `yarrr` to create plot
yarrr::pirateplot(
  formula = weight ~ Time,
  data = ChickWeight,
  theme = 1,
  main = stats_results
)

Usage and syntax simplicity

Overall consistency in API

Attempt has been made to make the application program interface (API) consistent enough that no struggle is expected while thinking about specifying function calls-

  • When a given function depends on variables in a dataframe, data argument must always be specified.
  • Often, package functions relevant for between-subjects versus within-subjects design expect tidy/long versus Cartesian/wide format data, respectively. ggstatsplot functions consistently expect tidy/long form data.
  • All functions accept both quoted (x = "var1") and unquoted (x = var1) arguments.
Consistent API

Consistent API

These set principles combined with the fact that almost all functions produce publication-ready plots that require very few arguments if one finds the aesthetic and statistical defaults satisfying make the syntax much less cognitively demanding and easy to remember/reconstruct.

Helpful messages and aesthetic modifications

ggstatsplot is a very chatty package and will by default print helpful notes on assumptions about statistical tests, warnings, etc. If you don’t want your console to be cluttered with such messages, they can be turned off by setting argument messages = FALSE in the function call.

All relevant functions in ggstatsplot have a return argument which can be used to not only return plots (which is the default), but also to return a subtitle or caption, which are objects of type call and can be used to display statistical details in conjunction with a custom plot and at a custom location in the plot.

Additionally, all functions share the ggtheme and palette arguments that can be used to specify your favorite ggplot theme and color palette.

Appendix

Appendix A: Documentation

There are three main documents one can rely on to learn how to use ggstatsplot:

Appendix B: Suggestions

If you find any bugs or have any suggestions/remarks, please file an issue on GitHub repository for this package: https://github.com/IndrajeetPatil/ggstatsplot/issues

Contributing

ggstatsplot is happy to receive bug reports, suggestions, questions, and (most of all) contributions to fix problems and add features. Pull Requests for contributions are encouraged.

Here are some simple ways in which one can contribute (in the increasing order of commitment):

  • Read and correct any inconsistencies in the documentation

  • Raise issues about bugs or wanted features

  • Review code

  • Add new functionality (in the form of new plotting functions or helpers for preparing subtitles)

Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.

Appendix D: Session information

For reproducibility purposes, the details about the session information in which this document was rendered, see- https://indrajeetpatil.github.io/ggstatsplot/articles/web_only/session_info.html

References

Nuijten, M. B., Hartgerink, C. H. J., van Assen, M. A. L. M., Epskamp, S., & Wicherts, J. M. (2016). The prevalence of statistical reporting errors in psychology (19852013). Behavior Research Methods, 48(4), 1205–1226. https://doi.org/10.3758/s13428-015-0664-2

  1. Harvard University, ↩

  2. Harvard University, ↩

  3. Harvard University, ↩